Add remove_from_index method

byaldi/RAGModel.py

@@ -155,6 +155,17 @@ def add_to_index(
             input_item, store_collection_with_index, doc_id, metadata=metadata
         )
 
+    def remove_from_index(self, doc_id: int):
+        """Remove an item to an existing index.
+
+        Parameters:
+            doc_id (int): The document ID for the item being removed.
+
+        Returns:
+            None
+        """
+        return self.model.remove_from_index(doc_id)
+
     def search(
         self,
         query: Union[str, List[str]],

byaldi/colpali.py

@@ -589,8 +589,62 @@ def _add_to_index(
         if self.verbose > 0:
             print(f"Added page {page_id} of document {doc_id} to index.")
 
-    def remove_from_index(self):
-        raise NotImplementedError("This method is not implemented yet.")
+    def remove_from_index(self, doc_id: int) -> Dict[int, str]:
+        if self.index_name is None:
+            raise ValueError(
+                "No index loaded. Use index() to create or load an index first."
+            )
+
+        if not hasattr(self, "doc_ids_to_file_names"):
+            raise ValueError("No documents in the index.")
+
+        if doc_id not in self.doc_ids_to_file_names:
+            raise ValueError(f"Document ID {doc_id} does not exist in the index.")
+
+        # Remove the document from doc_ids_to_file_names
+        del self.doc_ids_to_file_names[doc_id]
+
+        # Remove associated embeddings
+        embed_ids_to_remove = [
+            embed_id
+            for embed_id, doc_info in self.embed_id_to_doc_id.items()
+            if doc_info["doc_id"] == doc_id
+        ]
+        self.indexed_embeddings = [
+            embedding
+            for i, embedding in enumerate(self.indexed_embeddings)
+            if i not in embed_ids_to_remove
+        ]
+
+        # Create a new dictionary for embed_id_to_doc_id excluding entries in embed_ids_to_remove
+        remaining_embed_id_to_doc_id = {
+            embed_id: self.embed_id_to_doc_id[embed_id]
+            for embed_id in self.embed_id_to_doc_id
+            if embed_id not in embed_ids_to_remove
+        }
+
+        # Re-index the remaining items in embed_id_to_doc_id
+        self.embed_id_to_doc_id = {
+            new_id: doc_info
+            for new_id, doc_info in enumerate(remaining_embed_id_to_doc_id.values())
+        }
+
+        # Remove associated entries from the collection if they exist
+        if hasattr(self, "collection"):
+            for embed_id in embed_ids_to_remove:
+                if embed_id in self.collection:
+                    del self.collection[embed_id]
+
+        # Remove metadata if it exists
+        if hasattr(self, "doc_id_to_metadata") and doc_id in self.doc_id_to_metadata:
+            del self.doc_id_to_metadata[doc_id]
+
+        # Update the highest_doc_id if necessary
+        if doc_id == self.highest_doc_id:
+            self.highest_doc_id = max(self.doc_ids_to_file_names.keys(), default=-1)
+
+        self._export_index()
+        return self.doc_ids_to_file_names
 
     def filter_embeddings(self,filter_metadata:Dict[str,str]):
         req_doc_ids = []